Performance Evaluation of K-Means Clustering Using MapReduce Programming Model

In the current era of unparalleled data expansion, effective handling. large datasets have emerged as a crucial obstacle. When using enormous datasets that are terabytes or petabytes in size, the conventional k-means clustering approach has computational time limits. In the MapReduce framework, we assess the k-means algorithm.'s performance using multiple methods: K-means simple, K-means with Initial Equidistant Centres (IEC), and K-means Java implementation. on MapReduce. We investigate the newsgroup dataset and evaluate their performance in various infrastructures. settings. We also perform a comparative study at different iteration levels between the above-mentioned K-means methods. We use this study to demonstrate improvement in calculated time performance with various infrastructures. Additionally, we also analyze k means algorithms and their behavior with respect to centroids and various iteration levels, and hence provide deeper insights into their dynamics. Our paper offers useful benchmarks for further research and practices working with large-scale data clustering, illuminating the best methods to make use of parallel computation.